Transcriptional efficiency is impacted by both transcription factor (TF) missense variants and by variants that alter the DNA sequence in TF binding regions. For example, missense variants in the transcription factors TBR1 and NR4A2 are known to cause neurodevelopmental disorders, while those in ESRRB cause inherited forms of hearing loss. To understand the impact of TF missense variants and/or changes to cognate DNA binding regions, experimental methods such as SELEX sequencing have been successfully employed. As a complementary computational approach, molecular dynamics (MD)-based free-energy differences can be used to quantify changes in binding affinity as a function of either TF missense variations or changes to DNA base-pairs within binding regions. Current MD approaches have been limited to fixed charge force fields; however, emerging methods based on more advanced electrostatics are attractive given the highly charged nature of protein-nucleic acid interactions. The polarizable atomic multipole AMOEBA force field includes permanent atomic multipoles through quadrupole order and induced dipoles to capture electronic polarization. Although AMOEBA has shown promise for calculating protein-ligand absolute binding affinities that go beyond the accuracy of fixed charge force field models, support for GPU-accelerated dual-topology style alchemical relative free-energy simulations has not been available due to the challenges in completing an implementation. Here, we present for the first time AMOEBA dual-topology simulation results for computing relative free-energy differences that are fully accelerated on GPUs. We discuss solutions to implementation challenges using the Force Field X-OpenMM software, including work to optimize the efficiency of the alchemical path between end states and validation on DNA base pair substitutions. This, in turn, sets the stage for explaining how subtle missense variants cause neurodevelopmental disorders and inherited forms of hearing loss.
Miller et al. (Sun,) studied this question.